GPS Performance Compared : SiRAtlasIV Impresses

"When the SiRFstarIII chipset was released in early 2005, it
represented a major breakthrough for GPS as a mainstream consumer
product thanks to its high sensitivity engine that made GPS easier to
use. Since that time, GPS performance had remained unchanged, with
other chipset makers releasing their own high sensitivity chipsets,
see the 2007 GPS Chipsets reviewed article, and SiRF adding some "off
board" enhancements such as SiRFInstantFix AGPS version 1 and version
2. The announcement of SiRFstarIV in July has brought back some
excitement to the field and while we're waiting for it to be released,
here is a preview of things to come with the new SiRFAtlasIV PND
platform that uses a similar advanced engine."

http://www.gpspassion.com/fr/articles.asp?id=175&page=9

Honestly I don't understand why the use "Average # of satellites",
signal Level, fix validity and so on in the comparison...
They should know that NMEA data is not always true.
You can use DR to propagate validity and easily reach 100%, GSV CN0
indication and GSA satellites are also not the real things that are
happening inside the receiver.
And, also, using fewer (really, not in GSA) does not mean always to be
worst.

Moreover they seems to privilege track smothness instead of accuracy,
and the accuracy is never really measured as they look only at cross-
track accuracy. In the way they analyze the track, that totally ignore
the along-track accuracy.